[Bf-committers] Particle experiments - looking for feedback

Leon Turner bf-committers@blender.org
Wed, 5 May 2004 08:03:50 +0900

Hi Alex,

I was looking into this a bit yesterday: unfortunately the component
form stuff is not updated where I need it, so I would need to do an
matrix inverse operation.

So the trade off would be 1 one matrix inverse, and 2 point transforms
per object, versus the current system which does 4 matrix transforms
per face.

Any idea where the "break even" point should be: i.e. approximately
how many faces per object I would need before the performance would be
better than the current system?



On Mon, 03 May 2004 23:41:48 +0100, Alex Mole <nal@blueyonder.co.uk> wrote:
> Hi Leon
> It should take about the same amount of time to work out the inverse
> transform and multiply two particle points as it does to transform four
> vertices of a quad [roughly speaking]. However, once the particle has
> been inversely transformed, it will work for every face in the object.
> So for an object with 50 faces, we're talking 2 inverse transforms vs.
> 200 forward transforms. I'd say that'd be a speed improvement ;)
> I don't know enough about the guts of Blender to say for sure, but AFAIK
> transforms are stored in component {translation,rotation,scaling} form
> as well as a matrix, so you should be able to trivially construct the
> inverse transform.
> Again, feel free to ask for advice on this if you want it :)
> Keep up the good work!
> Alex
> Leon Turner wrote:
> > Alex - I wasn't sure whether transforming the particle co-ordinates /
> > velocities to object space would help much, given that I need to cycle
> > through many possible deflection objects. What do you think?
> _______________________________________________
> Bf-committers mailing list
> Bf-committers@blender.org
> http://www.blender.org/mailman/listinfo/bf-committers